Climate-related risks threaten lives and livelihoods along the globe. Floods, Droughts, Heatwaves, Tropical Storms, and Wildfires are among the most devastating natural hazards. Forecasts and Early Warning of Hazards and their induced impacts gives means to mitigate damages. Weather forecasts for fast-onset hazards such as floods and seasonal forecasts for slow-onset hazards such as drought are used in anticipatory action to protect lives and livelihood before the disaster strikes.  Beyond that time-scale preparedness and targeted longer term social and material investments can help reduce vulnerability, and thus risk. For a prioritization of measures an anticipation of longer-term vulnerabilities is important. We conjecture that recent developments in AI can help for both time-scales.

Through this workshop organized by the RISK KAN, ELLIS Unit Jena, German Red Cross and Croix Rouge Malagasy, we intended to explore if, how, and where Artificial Intelligence promised to fundamentally push climate risk mitigation. For this purpose, we established a dialogue between humanitarian practitioners and machine learning researchers. We discussed recent advances in AI for early warning, remaining challenges in the field, and possible ways forward to make AI accessible for the humanitarian community.